DeepMHADTA: Prediction of Drug-Target Binding Affinity Using Multi-Head Self-Attention and Convolutional Neural Network

Drug-target interactions provide insight into the drug-side effects and drug repositioning. However, wet-lab biochemical experiments are time-consuming and labor-intensive, and are insufficient to meet the pressing demand for drug research and development. With the rapid advancement of deep learning, computational methods are increasingly applied to screen drug-target interactions. Many methods consider this problem as a binary classification task (binding or not), but ignore the quantitative binding affinity. In this paper, we propose a new end-to-end deep learning method called DeepMHADTA, which uses the multi-head self-attention mechanism in a deep residual network to predict drug-target binding affinity. On two benchmark datasets, our method outperformed several current state-of-the-art methods in terms of multiple performance measures, including mean square error (MSE), consistency index (CI), rm2, and PR curve area (AUPR). The results demonstrated that our method achieved better performance in predicting the drug–target binding affinity.     

Rhizosphere soil bacterial communities and nitrogen cycling affected by deciduous and evergreen tree species

Deciduous and evergreen trees differ in their responses to drought and nitrogen (N) demand. Whether or not these functional types affect the role of the bacterial community in the N cycle during drought remains uncertain. Two deciduous tree species (Alnus cremastogyne, an N₂‐fixing species, and Liquidambar formosana) and two evergreen trees (Cunninghamia lanceolata and Pinus massoniana) were used to assess factors in controlling rhizosphere soil bacterial community and N cycling functions. Photosynthetic rates and biomass production of plants, 16S rRNA sequencing and N‐cycling‐related genes of rhizosphere soil were measured. The relative abundance of the phyla Actinobacteria and Firmicutes was higher, and that of Proteobacteria, Acidobacteria, and Gemmatimondaetes was lower in rhizosphere soil of deciduous trees than that of evergreen. Beta‐diversity of bacterial community also significantly differed between the two types of trees. Deciduous trees showed significantly higher net photosynthetic rates and biomass production than evergreen species both at well water condition and short‐term drought. Root biomass was the most important factor in driving soil bacterial community and N‐cycling functions than total biomass and aboveground biomass. Furthermore, 44 bacteria genera with a decreasing response and 46 taxa showed an increased response along the root biomass gradient. Regarding N‐cycle‐related functional genes, copy numbers of ammonia‐oxidizing bacteria (AOB) and autotrophic ammonia‐oxidizing archaea (AOA), N₂ fixation gene (nifH), and denitrification genes (nirK, nirS) were significantly higher in the soil of deciduous trees than in that of the evergreen. Structural equation models explained 50.2%, 47.6%, 48.6%, 49.4%, and 37.3% of the variability in copy numbers of nifH, AOB, AOA, nirK, and nirS, respectively, and revealed that root biomass had significant positive effects on copy numbers of all N‐cycle functional genes. In conclusion, root biomass played key roles in affecting bacterial community structure and soil N cycling. Our findings have important implications for our understanding of plants control over bacterial community and N‐cycling function in artificial forest ecosystems.     

抑郁症易感基因和抗抑郁药物靶点相关[]神经细胞类型及相互作用网络分析

基于抑郁症的全基因组关联分析研究(GWAS),对于获得的单核苷酸多态性位点(SNP)使用Haploreg软件进行基因注释,得到SNP注释的102个易感基因.。使用MAGMA软件对GWAS的汇总统计数据做基因水平的分析,获得了270个校正之后显著的基因,两者合并共得到320个抑郁症易感基因。通过药物数据库Drugbank获取133个抗抑郁药物靶点基因。使用EWCE包对抑郁症易感基因和抗抑郁药物靶点在三套脑组织单细胞测序数据中,分别进行神经细胞类型富集分析。结果发现大脑皮质的GABA神经元(抑制性神经元)和谷氨酸能神经元(兴奋性神经元)是抑郁症易感基因和抗抑郁药物靶点共同的神经元。这两种类型的神经细胞可能是抗抑郁药物与抑郁症易感基因相互作用的神经细胞,另外少突胶质前体细胞可能是抑郁症特有的易感神经细胞。使用Network Calculator软件构建网络并进行进行网络拓扑学参数分析。结果表明抑郁症易感基因与抗抑郁药物靶点组成了一个具有显著的相互连接的网络。本研究从单细胞层面揭示抑郁症的遗传机制,在网络层面为寻找新的抗抑郁药物靶点提供了一定的启示。     

DLGCN:基于图卷积网络的药物-lncRNA[]关联预测

为实现高通量识别新的药物-长链非编码RNA(Long non-coding RNA, lncRNA)关联,本文提出了一种基于图卷积网络模型来识别潜在药物-lncRNA关联的方法DLGCN(Drug-LncRNA graph convolution network)。首先,基于药物的结构信息和lncRNA的序列信息分别构建了药物-药物和lncRNA-lncRNA相似性网络,并整合实验证实的药物-lncRNA关联构建了药物-lncRNA异质性网络。然后,将注意力机制和图卷积运算应用于该网络中,学习药物和lncRNA的低维特征,基于整合的低维特征预测新的药物-lncRNA关联。通过效能评估,DLGCN的受试者工作特性曲线下面积(Area under receiver operating characteristic, AUROC)达到0.843 1,优于经典的机器学习方法和常见的深度学习方法。此外,DLGCN预测到姜黄素能够调控lncRNA MALAT1的表达,已被最近的研究证实。DLGCN能够有效预测药物-lncRNA关联,为肿瘤治疗新靶点的识别和抗癌药物的筛选提供了重要参考。     

Ecology,morphology and physiology of Chroothece richteriana (Rhodophyta,Stylonematophyceae) in the highly calcareous Río Chícamo,south-east Spain

The ecology of Chroothece was studied in the highly calcareous Río Chícamo, south-east Spain, in order to explain its success there, but rarity elsewhere. The river, which originates mainly from an underground aquifer, has water with high conductivity, sulphate and nitrate but low phosphate concentrations, the latter mainly organic. Chroothece occurs in mats and in lobed colonies reaching 4 cm in the broadest dimension. The colony surface consists of one layer of cells, each of which is attached to a stalk, which dichotomizes when the cell divides; stalks often extend to the colony base. The central region of many mat cells and almost all colony cells has a yellow to orange-brown colour, associated with the numerous lipid droplets densely covering the surface of the pyrenoid and arms of the star-shaped chloroplast. Field material and laboratory isolates indicate that stalk formation occurs under moderate P limitation and both stalks and cell sheath show high phosphatase activities. This also occurred in a culture collection strain maintained for 30 years in a very P-rich medium, but then transferred to a moderately P-limiting medium (c. 0.9 mg l^?1). We suggest that colony formation is initiated by aggregation of motile cells following P pulses in the water. Comparisons are made with Rivularia, a competitor in this nitrate-rich river, in spite of being a N₂-fixer. One difference is that Chroothece cells lie at the periphery of the colonies and are therefore exposed to maximum sunlight, whereas Rivularia trichomes grow inside colonies with photoprotection by scytonemin. However, the ability to withstand heavy grazing pressure may be an especially important factor favouring Rivularia here.     

基于小波神经网络的麦蚜发生程度预测模型

【目的】建立基于小波神经网络病虫害预测预报模型,对提前采取防病防虫措施、减少农作物病虫害损失、提高农作物产量与质量具有重要意义。【方法】本研究以山西省运城市芮城县1980-2014年麦蚜发生程度和气象因子数据为基础,采用主成分分析法从40个基础气象因子中整合形成9个新的自变量输入模型,采用试凑法筛选隐含层节点数,用1980-2009年的数据进行网络训练,对2010-2014年麦蚜发生程度进行回测,建立了以Morlet小波函数为传递函数的小波神经网络模型,并与以Sigmoid函数为传递函数的BP神经网络模型进行了比较。【结果】小波和BP神经网络两种模型对训练样本的平均拟合精度均有10年以上超过80%,两者MAPE 值分别为 89.83% 和83.07%,MSE 值分别为0.0578和0.6192。【结论】两个模型都能较好地描述麦蚜发生程度;从预测精度和模型的稳定性来看,小波神经网络好于BP神经网络。     

Changes in diatom-dominated biofilms during simulated improvements in water quality: implications for diatom-based monitoring in rivers

Although benthic diatoms are used to assess river water quality, there are few data on the rate at which diatom assemblages react to changes in water quality. The aim of this study was to assess the reaction time of diatoms and to discuss the changes occurring during water quality improvement on the basis of their autecological characteristics. In order to simulate this improvement, diatom-dominated biofilms grown on artificial sandstone substrata were transferred from several polluted rivers to an unpolluted river. They were sampled three times: before transfer and 1 and 2 months after transfer. The ecology and growth-forms of the taxa explained most of the changes in species composition observed during the experiment. Adnate diatoms gradually replaced motile and stalked taxa. Gomphonema parvulum, a stalked diatom positioned vertically in the biofilm, is adapted for light and space competition in high-density algal biofilms. When transferred to an unpolluted site, this growth-form is less competitive and does not tolerate the high grazing pressure. Fistulifera saprophila is a single celled motile diatom, living in organic matrices. When the artificial substrata were transferred to the unpolluted site, this particular ecological niche disappeared quickly. On the other hand, Achnanthidium minutissimum, which is considered to be cosmopolitan and an early colonizer, increased during the first month of transfer and then decreased. It was gradually replaced by A. biasolettianum, which was the taxon best suited to this pristine stream. The changes observed differed between treatments depending on the species composition and architecture of the biofilms. In particular, biofilms dominated by stalked and motile diatoms were more quickly modified than those dominated by small motile diatoms. The diatom index reflects these changes, and its values showed that about 60 days following a water quality improvement were necessary for transferred diatom assemblages to reach diatom index values similar as those at the unpolluted river.     

Organic matter processing by a stream-segment ecosystem: Fort River,Massachusetts, U.S.A.

An annual organic matter budget for a 1700 m segment of Fort River (Massachusetts, USA) is presented. Primary production in this fourth order stream exceeds litter input annually, however ecosystem P/R is 0.5. Respiration in excess of gross primary production is supported by allochthonous organic matter imported from upstream reaches. The relative contribution of organic matter size fractions to stream consumers depends upon biologic lability, rate of input, and residence time in the ecosystem. Particles of seston size (1 μm to 1 mm) are most heavily used by consumers, however dissolved organic matter represents the largest input component. Microorganisms are the predominant consumers in this soft-water, nutrient-poor stream ecosystem. A conceptual model for assessing the processing efficiency of stream ecosystems is presented and discussed in terms of several headwater to estuary gradients.